Research

Methodology notes and benchmark frameworks for evaluating multi-model document assurance—published without fabricated scores or undisclosed harness details.

Last updated 2026-05-14

Read the methodology before claims

Benchmark methodology
Fair evaluation design—no fabricated scores.
Benchmark framework
What a serious harness would measure.
Choosing tools
Procurement questions while research is in progress.
Editorial policy
How Academy and research pages are governed.

Topics

Benchmark methodology

Tasks, rubrics, adjudication, and disclosure rules before any leaderboard is credible.

How we think about evaluating document-assurance systems fairly.

Benchmark framework

Scope for future comparative work—explicitly without fabricated vendor scores today.

What would be measured when real runs exist.

If you are procuring or auditingPair research with product methodology.

Start with benchmark methodology, then read editorial policy. For workflow fit while evaluation is underway, see Choosing document assurance tools and product methodology.

Evaluate on real documents

Read the methodology, then run a Document Check when you are ready to test workflow fit.

Run a Document Check